The Trump administration has officially lifted the two-week ban on the export of Anthropic's Claude Mythos 5 model. This decision opens access to the advanced AI tool for over 100 American organizations, including major corporations and government agencies. This represents a strategic shift that brings Mythos 5 back into play, while the fate of its more powerful counterpart, Fable 5, remains uncertain.
In a letter to Anthropic CTO Tom Brown, Commerce Secretary Howard Lutnick explained that current precautionary measures allow the model to be transferred to trusted partners from Appendix A without requiring a separate license. Thus, the administration has acknowledged that the security risks that previously blocked exports are now considered manageable.
Mythos 5: From Secret Project to Broad Access
Recall that Mythos 5 was initially deployed exclusively within the Glasswing program, a vulnerability discovery project involving approximately 150 organizations from over 15 countries. Test results were impressive: the model identified critical errors in classified systems within hours. Now, the lifting of restrictions means that technology capable of finding "holes" in state cybersecurity will become available to a wide range of American partners.
A key trigger for the initial block was concern from Amazon, one of Anthropic's largest investors. Company researchers warned that Fable 5, unlike Mythos, could potentially be hacked for dangerous purposes. However, the situation with Mythos 5 is different: its vulnerability discovery capabilities were apparently deemed acceptable for expanded use.
Fable 5: Waiting and New Regulation
As for Fable 5, its release to a broad audience has been postponed. Previously, this model was available to all subscribers and was considered the most powerful public AI tool. Now, its release process has stalled, with no one specifying exact timelines. The situation is compounded by new regulation: on June 2, an executive order was signed establishing a voluntary channel for federal review of advanced AI models. Developers can submit their solutions for cybersecurity checks 30 days before launch.
This Washington move is part of a broader trend towards tightening control over AI technologies, reminiscent of restrictions on chip exports to China. OpenAI, incidentally, has followed a similar path, limiting access to the most powerful version of GPT-5.6, named Sol, to just 20 state-approved partners.
Global Context and My View
It is worth noting that European and other countries have become heavily dependent on Washington's decisions regarding access to new developments. A letter organized by former Facebook security chief Alex Stamos and signed by leaders of Nvidia, Adobe, Zoom, and other companies urged authorities to abandon restrictive measures. However, as we see, the administration has chosen a path of controlled access.
My opinion: Lifting the ban on Mythos 5 is a pragmatic step that will strengthen the cybersecurity of American structures, but it also sets a dangerous precedent. Now, control over the distribution of the most powerful AI models will depend entirely on political will, rather than market mechanisms. Fable 5 will likely face a similar fate, but with even stricter access conditions. The AI market is entering an era of "managed scarcity," where access to cutting-edge technology will become a privilege, not a right.